review

name	review
description	Run a structured code review on recent changes or specified files
user-invocable	true

Perform a structured code review. If arguments are provided, review those files/paths. If no arguments, review recent uncommitted or staged changes.

Review Checklist

Work through each category and report findings.

1. Correctness

• Does the code do what it claims?
• Are edge cases handled?
• Are error paths meaningful (not swallowed or generic)?
• Are types accurate and strict (no unnecessary any)?

2. Design Doc Compliance

Cross-reference changes against docs/dev/ (architecture.md, schema-reference.md, etc.):

• Does the data model match confirmed schema?
• Does CLI surface match the intended commands?
• Are MCP guardrails respected (structured tools for metadata, no direct frontmatter editing)?
• Are workflow values using config keys not labels?
• Is the .loctt/ directory structure correct?

Flag anything that contradicts a confirmed decision.

3. Clean Code (per .claude/housekeeping.md)

• No debug logging left behind
• No dead code or commented-out blocks
• Comments describe WHY not WHAT
• No quick fixes or workarounds
• No any types without justification
• Named exports used

4. Impact Analysis

• What else in the codebase is affected by this change?
• Are there callers, tests, or docs that need updating?
• Any breaking changes to existing behavior?

5. Testing Gaps

• What's untested that should be?
• Are mocks limited to external dependencies only?
• Would existing tests catch a regression if this code broke?

Output Format

## Review: [files or scope]

### Correctness
- [findings]

### Design Doc Compliance
- [findings or "No issues"]

### Clean Code
- [findings]

### Impact
- [affected areas]

### Testing Gaps
- [what needs tests]

### Summary
[severity] [count] issues found
- Critical: [list]
- Suggestions: [list]

Arguments: $ARGUMENTS